Classify each listing photo as ready to publish, needs work, or CGI render.
The product was shipping 50,000 photos per month to a frontier vision model, costing about $3,000 per month ($36,000 per year). After the audit, the same task ran on a smaller, cheaper vision model with comparable accuracy, dropping the bill to roughly $200 per month ($2,400 per year). Net annual savings: $33,600. We also delivered a fallback ladder for the harder edge cases.
~15x
cheaper per call
Same
accuracy at human review
2.4x
faster per request
$36,000 / yr
before audit
$2,400 / yr
after audit
Edge cases the audit specifically planned for
- Subtle CGI renders that look photographic at first glance
- Mixed listings (some real photos, some renders in the same gallery)
- Light-work vs major-work properties (fresh paint vs gut renovation)
- Construction-progress photos vs finished interiors
- Lighting, photo quality, and angle variance across providers
Vision tasks are easy to picture, but the same playbook applies to text classification, document extraction, ticket routing, RAG grading, content moderation, and more. Edge-case planning is the audit's value-add. Without it, score variance across models collapses and recommendations become unreliable.